Recent developments in MEMS and smart structures combined with the low price of microprocessors call for new methods of distributed control for large systems.
Our approach to distributed control involves diagonalizing the plant matrix, or decoupling the interactions between the inputs and outputs.
SVD factors exactly diagonalize a real plant matrix, require global information, and are not scalable.
We have proposed a method of approximately diagonalizing a symmetric plant matrix. This method
- requires only nearest neighbor communication
- naturally distributes the computational burden
- can be found by numerically integrating a system of coupled ODEs